Skip to main content
European Commission logo print header

A Pragmatic-Based Language Understanding System

Objetivo

PLUS achieves robustness in natural language understanding by treating natural language as a communicative activity whose essential characteristic is to convey a meaning that is both appropriate and contextually relevant.
An environment for developing intelligent natural language interfaces to interactive computing environments (both information seeking dialogue systems and problem solving environments) has been developed. A first demonstrator, interactive Yellow Pages, was developed; the modular architecture incorporates a replaceable application model, enabling new applications to be produced with relatively little effort.

Currently available languages are English and French. The emphasis on pragmatics and language independent reasoning, and the resulting simplicity and modularity of the language specific knowledge bases, facilitates the adaptation of the system to new languages or multilingual use.
Since the intention of a human user of natural language is to convey a message, and since all messages occur in some context, it is crucially important to exploit this context in the derivation of the intended interpretation. This contrasts with other approaches to the problem of robustness, which miss the heart of the problem, which is to react appropriately in a context created or updated by the fact that a user has typed something at a keyboard with the express purpose of communicating a message.

One key issue is the exploitation of both pragmatic and linguistic phenomena (such as interpretation with respect to context) in order to provide reasoning-based robustness in natural language understanding by integrating these two areas.

Tema(s)

Data not available

Convocatoria de propuestas

Data not available

Régimen de financiación

Data not available

Coordinador

CAP Gemini Innovation
Aportación de la UE
Sin datos
Dirección
7 chemin du Vieux Chêne
38240 Meylan
Francia

Ver en el mapa

Coste total
Sin datos

Participantes (7)